| any suggestion about merging small regions [message #30571] |
Tue, 07 May 2002 22:25  |
Xiaoying Jin
Messages: 20
Registered: May 2002
|
Junior Member |
|
|
Hi, there,
After segmentation, I got a lot of regions contiguous with each other.
As Ted Cary descibed in the previous post,
"It is something like a grayscale photograph of a jar of
marbles. All the marbles are touching each other, so there is no
background. Two marbles of the same color may not belong to the same
region.
You want to give a unique label to each marble/region."
But now here is the problem: Even if the gray levels of the regions are
fairly uniform,
there are a lot of holes in a big region assigned to other different
regions. So if I try to
find edges between different regions, the output will get complicated.
To solve this problem, I want to merge regions with small area into the
adjacent regions.
Any suggestion about how to implement it in IDL?
Best regards,
Xiaoying Jin
|
|
|
|
|
|
|
|
| Re: any suggestion about merging small regions [message #30695 is a reply to message #30571] |
Thu, 09 May 2002 09:20  |
Ted Cary
Messages: 53
Registered: October 2001
|
Member |
|
|
"Ted Cary" <tedcary@yahoo.com> wrote in message
news:abe81u$ggg$1@netnews.upenn.edu...
> MORPH_CLOSE did not work for this? Out of curiosity, did you apply it to
> your original image or to the bi-level region mask?
Sorry, I think I should have been recommending MORPH_OPEN instead of
MORPH_CLOSE. Hopefully, you figured it out.
Fanning, keep your coyote on a leash...
TC
|
|
|
|
| Re: any suggestion about merging small regions [message #30696 is a reply to message #30571] |
Thu, 09 May 2002 08:29 |
Ted Cary
Messages: 53
Registered: October 2001
|
Member |
|
|
"Xiaoying Jin" <xje4e@mizzou.edu> wrote in message
news:abacij$lgn$1@dipsy.missouri.edu...
>
> But now here is the problem: Even if the gray levels of the regions are
> fairly uniform,
> there are a lot of holes in a big region assigned to other different
> regions. So if I try to
> find edges between different regions, the output will get complicated.
> To solve this problem, I want to merge regions with small area into the
> adjacent regions.
>
MORPH_CLOSE did not work for this? Out of curiosity, did you apply it to
your original image or to the bi-level region mask?
You might try to SMOOTH the image, or you can reduce the colors in it. You
mentioned this is an unsigned integer image; that's a lot of gray levels, do
you need them all? If the regions are truly almost uniform in gray level,
reducing the number of grays might get rid of the noise.
TC
|
|
|
|
| Re: any suggestion about merging small regions [message #30711 is a reply to message #30571] |
Wed, 08 May 2002 10:24 |
Xiaoying Jin
Messages: 20
Registered: May 2002
|
Junior Member |
|
|
"Ted Cary" <tedcary@yahoo.com> wrote in message
news:abb87e$5uo$1@netnews.upenn.edu...
> How about MORPH_CLOSE?
>
> TC
I tried that, it helps, but not good enough. If I choose the small structure
element, there is still lots of small
regions left. If I choose the large structure element, it will merge some
near large regions together and
affects the shape of the regions.
Regards,
Xiaoying Jin
|
|
|
|
| Re: any suggestion about merging small regions [message #30715 is a reply to message #30571] |
Wed, 08 May 2002 05:14 |
Ted Cary
Messages: 53
Registered: October 2001
|
Member |
|
|
How about MORPH_CLOSE?
TC
"Xiaoying Jin" <xje4e@mizzou.edu> wrote in message
news:abacij$lgn$1@dipsy.missouri.edu...
> Hi, there,
>
> After segmentation, I got a lot of regions contiguous with each other.
>
> As Ted Cary descibed in the previous post,
> "It is something like a grayscale photograph of a jar of
> marbles. All the marbles are touching each other, so there is no
> background. Two marbles of the same color may not belong to the same
> region.
> You want to give a unique label to each marble/region."
>
> But now here is the problem: Even if the gray levels of the regions are
> fairly uniform,
> there are a lot of holes in a big region assigned to other different
> regions. So if I try to
> find edges between different regions, the output will get complicated.
> To solve this problem, I want to merge regions with small area into the
> adjacent regions.
>
> Any suggestion about how to implement it in IDL?
>
> Best regards,
>
> Xiaoying Jin
>
>
>
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
